/* Flite_HTS_Engine_synthesize: synthesize speech */
HTS_Boolean Flite_HTS_Engine_synthesize(Flite_HTS_Engine * f, const char *txt, const char *wav)
{
int i;
FILE *fp;
cst_voice *v = NULL;
cst_utterance *u = NULL;
cst_item *s = NULL;
char **label_data = NULL;
int label_size = 0;
if (txt == NULL)
return FALSE;
/* text analysis part */
v = REGISTER_VOX(NULL);
if (v == NULL)
return FALSE;
u = flite_synth_text(txt, v);
if (u == NULL) {
UNREGISTER_VOX(v);
return FALSE;
}
for (s = relation_head(utt_relation(u, "Segment")); s; s = item_next(s))
label_size++;
if (label_size <= 0) {
delete_utterance(u);
UNREGISTER_VOX(v);
return FALSE;
}
label_data = (char **) calloc(label_size, sizeof(char *));
for (i = 0, s = relation_head(utt_relation(u, "Segment")); s; s = item_next(s), i++) {
label_data[i] = (char *) calloc(MAXBUFLEN, sizeof(char));
Flite_HTS_Engine_create_label(f, s, label_data[i]);
}
/* speech synthesis part */
HTS_Engine_synthesize_from_strings(&f->engine, label_data, label_size);
if (wav != NULL) {
fp = fopen(wav, "wb");
HTS_Engine_save_riff(&f->engine, fp);
fclose(fp);
}
HTS_Engine_refresh(&f->engine);
for (i = 0; i < label_size; i++)
free(label_data[i]);
free(label_data);
delete_utterance(u);
UNREGISTER_VOX(v);
return TRUE;
}
void test_hrg(void)
{
cst_utterance *u;
cst_relation *r;
cst_item *item = 0;
int i;
u = new_utterance();
r = utt_relation_create(u, "Segment");
for (i = 0; i < 10; i++)
{
char buff[20];
sprintf(buff, "seg_%03d", i);
if (i == 0)
item = relation_append(r, NULL);
else
item = item_append(item, NULL);
item_set_string(item, "name", buff);
item_set_float(item, "duration", i * 0.20);
}
for (i = 0, item = relation_head(utt_relation(u, "Segment"));
item; item = item_next(item), i++)
{
TEST_CHECK(item_feat_float(item, "duration") == correct_list[i]);
}
delete_utterance(u);
}
int main(int argc, char **argv)
{
cst_utterance *u;
cst_relation *r;
cst_item *item=0;
int i;
u = new_utterance();
r = utt_relation_create(u,"Segment");
for (i=0; i<10; i++)
{
char buff[20];
sprintf(buff,"seg_%03d",i);
if (i==0)
item = relation_append(r,NULL);
else
item = item_append(item,NULL);
item_set_string(item,"name",buff);
item_set_float(item,"duration",i*0.20);
}
for (i=0,item=relation_head(utt_relation(u,"Segment"));
item; item=item_next(item),i++)
{
printf("Segment %d %s %f\n",
i,
item_feat_string(item,"name"),
item_feat_float(item,"duration"));
}
delete_utterance(u);
return 0;
}
int main(int argc, char **argv)
{
cst_utterance *u;
cst_relation *r;
cst_item *item=0;
int i;
cmu_lex_init();
u = new_utterance();
r = utt_relation_create(u,"Word");
bbb_relation_load(r,"ttt.txt");
WordSylSeg(u);
for (i=0,item=item_next(relation_head(utt_relation(u,"Segment")));
item; item=item_next(item),i++)
{
printf("Segment %s %s %s %s\n",
ffeature_string(item,"name"),
ffeature_string(item,"n.name"),
ffeature_string(item,"p.name"),
ffeature_string(item,"R:SylStructure.parent.name")
/* ffeature_string(item,"R:SylStructure.parent.R:Word.n.name"), */
/* item_feat_float(item,"duration")); */
);
}
delete_utterance(u);
return 0;
}
cst_wave *flite_text_to_wave(const char *text, cst_voice *voice)
{
cst_utterance *u;
cst_wave *w;
if ((u = flite_synth_text(text,voice)) == NULL)
return NULL;
w = copy_wave(utt_wave(u));
delete_utterance(u);
return w;
}
float flite_phones_to_speech(const char *text,
cst_voice *voice,
const char *outtype)
{
cst_utterance *u;
float dur;
u = flite_synth_phones(text,voice);
dur = flite_process_output(u,outtype,FALSE);
delete_utterance(u);
return dur;
}
static cst_utterance *flite_synth_foo(cst_utterance *u,
cst_voice *voice,
cst_uttfunc synth)
{
utt_init(u, voice);
if ((*synth)(u) == NULL)
{
delete_utterance(u);
return NULL;
}
else
return u;
}
/* Flite_Text_Analyzer_analysis: text analysis */
void Flite_Text_Analyzer_analysis(Flite_Text_Analyzer * analyzer, const char *text)
{
int i;
cst_item *s;
Flite_Utterance *fu;
if (analyzer == NULL || text == NULL)
return;
if (analyzer->pointer != NULL)
Flite_Text_Analyzer_clear(analyzer);
/* allocate */
fu = (Flite_Utterance *) malloc(sizeof(Flite_Utterance));
/* create voice */
fu->v = REGISTER_VOX(NULL);
if (fu->v == NULL) {
free(fu);
return;
}
/* create utterance */
fu->u = flite_synth_text(text, fu->v);
if (fu->u == NULL) {
UNREGISTER_VOX(fu->v);
free(fu);
return;
}
/* count number of phonemes */
for (fu->nitem = 0, s = relation_head(utt_relation(fu->u, "Segment")); s; s = item_next(s), fu->nitem++);
if (fu->nitem == 0) {
delete_utterance(fu->u);
UNREGISTER_VOX(fu->v);
free(fu);
return;
}
/* save informations */
fu->items = (cst_item **) malloc(sizeof(cst_item *) * fu->nitem);
for (i = 0, s = relation_head(utt_relation(fu->u, "Segment")); s; s = item_next(s), i++)
fu->items[i] = s;
analyzer->pointer = (void *) fu;
}
/* Flite_Text_Analyzer_clear: finalize flite front-end */
void Flite_Text_Analyzer_clear(Flite_Text_Analyzer * analyzer)
{
Flite_Utterance *fu;
if (analyzer == NULL || analyzer->pointer == NULL)
return;
fu = (Flite_Utterance *) analyzer->pointer;
if (fu->items != NULL)
free(fu->items);
if (fu->u != NULL)
delete_utterance(fu->u);
if (fu->v != NULL)
UNREGISTER_VOX(fu->v);
free(fu);
analyzer->pointer = NULL;
}
/* Flite_HTS_Engine_synthesis: speech synthesis */
void Flite_HTS_Engine_synthesis(Flite_HTS_Engine * f, char *txt, FILE * wavfp)
{
int i;
cst_voice *v = NULL;
cst_utterance *u = NULL;
cst_item *s = NULL;
char **label_data = NULL;
int label_size = 0;
/* text analysis part */
v = REGISTER_VOX(NULL);
if (v == NULL)
return;
u = flite_synth_text(txt, v);
if (u == NULL)
return;
for (s = relation_head(utt_relation(u, "Segment")); s; s = item_next(s))
label_size++;
if (label_size <= 0)
return;
label_data = (char **) calloc(label_size, sizeof(char *));
for (i = 0, s = relation_head(utt_relation(u, "Segment")); s;
s = item_next(s), i++) {
label_data[i] = (char *) calloc(MAXBUFLEN, sizeof(char));
Flite_HTS_Engine_create_label(f, s, label_data[i]);
}
/* speech synthesis part */
HTS_Engine_load_label_from_string_list(&f->engine, label_data, label_size);
HTS_Engine_create_sstream(&f->engine);
HTS_Engine_create_pstream(&f->engine);
HTS_Engine_create_gstream(&f->engine);
if (wavfp != NULL)
HTS_Engine_save_riff(&f->engine, wavfp);
HTS_Engine_refresh(&f->engine);
for (i = 0; i < label_size; i++)
free(label_data[i]);
free(label_data);
delete_utterance(u);
UNREGISTER_VOX(v);
}
float flite_tokens_to_speech(cst_utterance *u,
cst_voice *voice,
const char *outtype)
{
cst_wave *w;
float durs;
u = flite_synth_foo(u,voice,utt_synth_tokens);
if (u == NULL)
return -1;
w = utt_wave(u);
durs = (float)w->num_samples/(float)w->sample_rate;
if (cst_streq(outtype,"play"))
play_wave(w);
else if (!cst_streq(outtype,"none"))
cst_wave_append_riff(w,outtype);
delete_utterance(u);
return durs;
}
string getPhonemes( const char* sText )
{
string sRet;
cst_features* args = new_features();
cst_voice* v;
cst_utterance* u;
cst_item* s;
const char* name;
//const cst_val* d;
flite_init();
v = register_cmu_us_no_wave( NULL );
u = flite_synth_text( sText, v );
for ( s = relation_head( utt_relation( u, "Segment" ) ); s; s = item_next( s ) )
{
sRet += item_feat_string( s, "name" );
float test = item_feat_float( s, "end" );
//d = segment_duration( s );
/* If its a vowel and is stressed output stress value */
if ( ( cst_streq( "+", ffeature_string( s, "ph_vc" ) ) ) &&
( cst_streq( "1", ffeature_string( s, "R:SylStructure.parent.stress" ) ) ) )
{
sRet += "1";
}
sRet += " ";
}
delete_utterance( u );
delete_features( args );
return sRet;
}
cst_utterance *flowm_utt_callback(cst_utterance *u)
{
char rst[FL_MAX_MSG_CHARS];
const char *tok;
cst_item *item;
char *space;
int extend_length;
/* In order to stop the synthesizer if the STOP button is pressed */
/* This stops the synthesis of the next utterance */
if ((flowm_play_status == FLOWM_PLAY) ||
(flowm_play_status == FLOWM_SKIP))
{
if (TTSWindow)
{
rst[0] = '\0';
space = "";
for (item=relation_head(utt_relation(u,"Token"));
item; item=item_next(item))
{
tok = item_feat_string(item,"name");
if (cst_streq("",space))
/* Only do this on the first token/word */
flowm_file_pos = item_feat_int(item,"file_pos");
extend_length = cst_strlen(rst) + 1 +
cst_strlen(item_feat_string(item,"prepunctuation"))+
cst_strlen(item_feat_string(item,"punc"));
if (cst_strlen(tok)+extend_length+4 < FL_MAX_MSG_CHARS)
cst_sprintf(rst,"%s%s%s%s%s",rst,space,
item_feat_string(item,"prepunctuation"),
tok,
item_feat_string(item,"punc"));
else
{
if (cst_strlen(rst)+4 < FL_MAX_MSG_CHARS)
cst_sprintf(rst,"%s ...",rst);
break;
}
space = " ";
}
if (flowm_file_pos > flowm_prev_utt_pos[flowm_utt_pos_pos])
{
if ((flowm_utt_pos_pos+1) >= FLOWM_NUM_UTT_POS)
{
/* Filled it up, so move it down */
memmove(flowm_prev_utt_pos,&flowm_prev_utt_pos[1],
sizeof(int)*(FLOWM_NUM_UTT_POS-10));
flowm_utt_pos_pos = (FLOWM_NUM_UTT_POS-10);
}
flowm_utt_pos_pos++;
flowm_prev_utt_pos[flowm_utt_pos_pos] = flowm_file_pos;
}
/* Send text to TTSWindow */
mbstowcs(fl_tts_msg,rst,FL_MAX_MSG_CHARS);
SetDlgItemText(TTSWindow, FL_SYNTHTEXT, fl_tts_msg);
/* Update file pos percentage in FilePos window */
cst_sprintf(rst,"%2.3f",flowm_find_file_percentage());
mbstowcs(fl_fp_msg,rst,FL_MAX_MSG_CHARS);
SetDlgItemText(TTSWindow, FL_FILEPOS, fl_fp_msg);
SystemIdleTimerReset(); /* keep alive while synthesizing */
if (flowm_play_status == FLOWM_SKIP)
flowm_play_status = FLOWM_PLAY;
}
return u;
}
else
{
delete_utterance(u);
return 0;
}
}
float flite_text_to_speech_phenome( const char* text,
cst_voice* voice,
const char* outtype,
void* pStream )
{
cst_utterance* u;
float dur;
float end_last = 0;
float end_current = 0;
float dur_current = 0;
float dur_sum = 0;
//feat_set_float( voice->features, "duration_stretch", 1 );
u = flite_synth_text( text, voice );
cst_item* s;
string sRet;
int nPhoneme = 0;
for ( s = relation_head( utt_relation( u, "Segment" ) ); s; s = item_next( s ) )
{
SPhenomeTiming ps;
string sPhoneme = item_feat_string( s, "name" );
sRet += sPhoneme;
end_current = item_feat_float( s, "end" );
dur_current = end_current - end_last;
//if ( !( nPhoneme == 0 && sPhoneme == "pau" ) )
//{
dur_sum += dur_current;
//}
ps.fWeight = 1;
/* If its a vowel and is stressed output stress value */
if ( ( cst_streq( "+", ffeature_string( s, "ph_vc" ) ) ) &&
( cst_streq( "1", ffeature_string( s, "R:SylStructure.parent.stress" ) ) ) )
{
sRet += "1";
ps.fWeight = 1.3;
}
sRet += " ";
if ( pStream )
{
// fade into each other
ps.sName = sPhoneme;
ps.fStart = end_current - dur_current;
ps.fEnd = end_current;
ps.fDuration = dur_current;
( ( CryMT::queue<SPhenomeTiming>* )pStream )->push( ps );
}
end_last = end_current;
++nPhoneme;
}
dur = flite_process_output( u, outtype, FALSE );
delete_utterance( u );
return dur;
}
float flite_file_to_speech(const char *filename,
cst_voice *voice,
const char *outtype)
{
cst_utterance *utt;
cst_tokenstream *ts;
const char *token;
cst_item *t;
cst_relation *tokrel;
float durs = 0;
int num_tokens;
cst_wave *w;
cst_breakfunc breakfunc = default_utt_break;
cst_uttfunc utt_user_callback = 0;
int fp;
if ((ts = ts_open(filename,
get_param_string(voice->features,"text_whitespace",NULL),
get_param_string(voice->features,"text_singlecharsymbols",NULL),
get_param_string(voice->features,"text_prepunctuation",NULL),
get_param_string(voice->features,"text_postpunctuation",NULL)))
== NULL)
{
cst_errmsg("failed to open file \"%s\" for reading\n",
filename);
return 1;
}
fp = get_param_int(voice->features,"file_start_position",0);
if (fp > 0)
ts_set_stream_pos(ts,fp);
if (feat_present(voice->features,"utt_break"))
breakfunc = val_breakfunc(feat_val(voice->features,"utt_break"));
if (feat_present(voice->features,"utt_user_callback"))
utt_user_callback = val_uttfunc(feat_val(voice->features,"utt_user_callback"));
/* If its a file to write to, create and save an empty wave file */
/* as we are going to incrementally append to it */
if (!cst_streq(outtype,"play") &&
!cst_streq(outtype,"none") &&
!cst_streq(outtype,"stream"))
{
w = new_wave();
cst_wave_resize(w,0,1);
cst_wave_set_sample_rate(w,16000);
cst_wave_save_riff(w,outtype); /* an empty wave */
delete_wave(w);
}
num_tokens = 0;
utt = new_utterance();
tokrel = utt_relation_create(utt, "Token");
while (!ts_eof(ts) || num_tokens > 0)
{
token = ts_get(ts);
if ((cst_strlen(token) == 0) ||
(num_tokens > 500) || /* need an upper bound */
(relation_head(tokrel) &&
breakfunc(ts,token,tokrel)))
{
/* An end of utt, so synthesize it */
if (utt_user_callback)
utt = (utt_user_callback)(utt);
if (utt)
{
utt = flite_do_synth(utt,voice,utt_synth_tokens);
durs += flite_process_output(utt,outtype,TRUE);
delete_utterance(utt); utt = NULL;
}
else
break;
if (ts_eof(ts)) break;
utt = new_utterance();
tokrel = utt_relation_create(utt, "Token");
num_tokens = 0;
}
num_tokens++;
t = relation_append(tokrel, NULL);
item_set_string(t,"name",token);
item_set_string(t,"whitespace",ts->whitespace);
item_set_string(t,"prepunctuation",ts->prepunctuation);
item_set_string(t,"punc",ts->postpunctuation);
/* Mark it at the beginning of the token */
item_set_int(t,"file_pos",
ts->file_pos-(1+ /* as we are already on the next char */
cst_strlen(token)+
cst_strlen(ts->prepunctuation)+
cst_strlen(ts->postpunctuation)));
item_set_int(t,"line_number",ts->line_number);
}
delete_utterance(utt);
ts_close(ts);
return durs;
}
float flite_file_to_speech(const char *filename,
cst_voice *voice,
const char *outtype)
{
cst_utterance *utt;
cst_tokenstream *ts;
const char *token;
cst_item *t;
cst_relation *tokrel;
float d, durs = 0;
int num_tokens;
cst_breakfunc breakfunc = default_utt_break;
if ((ts = ts_open(filename,
get_param_string(voice->features,"text_whitespace",NULL),
get_param_string(voice->features,"text_singlecharsymbols",NULL),
get_param_string(voice->features,"text_prepunctuation",NULL),
get_param_string(voice->features,"text_postpunctuation",NULL)))
== NULL)
{
cst_errmsg("failed to open file \"%s\" for reading\n",
filename);
return 1;
}
if (feat_present(voice->features,"utt_break"))
breakfunc = val_breakfunc(feat_val(voice->features,"utt_break"));
/* If its a file to write to delete it as we're going to */
/* incrementally append to it */
if (!cst_streq(outtype,"play") && !cst_streq(outtype,"none"))
{
cst_wave *w;
w = new_wave();
cst_wave_resize(w,0,1);
cst_wave_set_sample_rate(w,16000);
cst_wave_save_riff(w,outtype); /* an empty wave */
delete_wave(w);
}
num_tokens = 0;
utt = new_utterance();
tokrel = utt_relation_create(utt, "Token");
while (!ts_eof(ts) || num_tokens > 0)
{
token = ts_get(ts);
if ((strlen(token) == 0) ||
(num_tokens > 500) || /* need an upper bound */
(relation_head(tokrel) &&
breakfunc(ts,token,tokrel)))
{
/* An end of utt */
d = flite_tokens_to_speech(utt,voice,outtype);
utt = NULL;
if (d < 0)
goto out;
durs += d;
if (ts_eof(ts))
goto out;
utt = new_utterance();
tokrel = utt_relation_create(utt, "Token");
num_tokens = 0;
}
num_tokens++;
t = relation_append(tokrel, NULL);
item_set_string(t,"name",token);
item_set_string(t,"whitespace",ts->whitespace);
item_set_string(t,"prepunctuation",ts->prepunctuation);
item_set_string(t,"punc",ts->postpunctuation);
item_set_int(t,"file_pos",ts->file_pos);
item_set_int(t,"line_number",ts->line_number);
}
out:
delete_utterance(utt);
ts_close(ts);
return durs;
}
static float flite_ssml_to_speech_ts(cst_tokenstream *ts,
cst_voice *voice,
const char *outtype)
{
cst_features *ssml_feats, *ssml_word_feats;
cst_features *attributes;
const char *token;
char *tag;
cst_utterance *utt;
cst_relation *tokrel;
int num_tokens;
cst_breakfunc breakfunc = default_utt_break;
cst_uttfunc utt_user_callback = 0;
float durs = 0.0;
cst_item *t;
ssml_feats = new_features();
ssml_word_feats = new_features();
set_charclasses(ts,
" \t\n\r",
ssml_singlecharsymbols_general,
get_param_string(voice->features,"text_prepunctuation",""),
get_param_string(voice->features,"text_postpunctuation","")
);
if (feat_present(voice->features,"utt_break"))
breakfunc = val_breakfunc(feat_val(voice->features,"utt_break"));
if (feat_present(voice->features,"utt_user_callback"))
utt_user_callback = val_uttfunc(feat_val(voice->features,"utt_user_callback"));
num_tokens = 0;
utt = new_utterance();
tokrel = utt_relation_create(utt, "Token");
while (!ts_eof(ts) || num_tokens > 0)
{
token = ts_get(ts);
if (cst_streq("<",token))
{ /* A tag */
tag = cst_upcase(ts_get(ts));
if (cst_streq("/",tag)) /* an end tag */
{
tag = cst_upcase(ts_get(ts));
attributes = ssml_get_attributes(ts);
feat_set_string(attributes,"_type","end");
}
else
attributes = ssml_get_attributes(ts);
utt = ssml_apply_tag(tag,attributes,utt,ssml_word_feats);
cst_free(tag);
}
else if (cst_streq("&",token))
{ /* an escape sequence */
/* skip to ; and insert value in rawdata */
}
else
{
if ((cst_strlen(token) == 0) ||
(num_tokens > 500) || /* need an upper bound */
(relation_head(tokrel) &&
breakfunc(ts,token,tokrel)))
{
/* An end of utt, so synthesize it */
if (utt_user_callback)
utt = (utt_user_callback)(utt);
if (utt)
{
utt = flite_do_synth(utt,voice,utt_synth_tokens);
durs += flite_process_output(utt,outtype,TRUE);
delete_utterance(utt); utt = NULL;
}
else
break;
if (ts_eof(ts)) break;
utt = new_utterance();
tokrel = utt_relation_create(utt, "Token");
num_tokens = 0;
}
num_tokens++;
t = relation_append(tokrel, NULL);
item_set_string(t,"name",token);
item_set_string(t,"whitespace",ts->whitespace);
item_set_string(t,"prepunctuation",ts->prepunctuation);
item_set_string(t,"punc",ts->postpunctuation);
/* Mark it at the beginning of the token */
item_set_int(t,"file_pos",
ts->file_pos-(1+ /* as we are already on the next char */
cst_strlen(token)+
cst_strlen(ts->prepunctuation)+
cst_strlen(ts->postpunctuation)));
item_set_int(t,"line_number",ts->line_number);
}
}
delete_utterance(utt);
return durs;
}
